Database Reference
In-Depth Information
Category
Median/mean/most typical
Name
Simple typicality # games Min. per game
median
R. Gomes
0
.
0271
61
22
.
6
All players
mean
N/A
0
.
0307
54
.
4
20
.
51
most typical
D. Granger
0
.
3830
78
22
.
6
0
.
0903
median
J. Voskuhl
51
16
Centers
mean
N/A
0
.
0679
52
.
42
17
.
36
most typical
F. Elson
0
.
1041
72
21
.
9
median
A. Jefferson
0
.
0747
59
18
Forwards
mean
N/A
0
.
0509
54
.
83
19
.
97
most typical
M. Taylor
0
.
0910
67
18
.
1
median
C. Bell
0
.
0488
59
21
.
7
Guards
mean
N/A
0
.
0230
54
.
54
21
.
73
most typical
R. Murray
0
.
0756
76
27
.
8
Table 4.6
Comparison among medians, means, and typical players in the NBA data set.
1000
18
100
16
14
10
12
1
10
UB(LT3)
UB(DLTA)
LT3
DLTA
0.1
6
8
RT
LT3
DLTA
0.01
0
2
4
0.001
0.0001
2
2.5
3
3.5
4
5
10
15
20
25
Neighborhood threshold
Parameter k
(a) Error rate vs. neighborhood.
(b) Error rate vs.
k
.
12
30
RT
LT3
DLTA
10
25
8
20
6
15
4
RT
LT3
DLTA
10
2
5
0
0
5
10
15
20
25
10000
15000
20000
25000
30000
Number of attributes
Number of tuples
(c) Error rate vs. dimensionality.
(d) Error rate vs. cardinality.
Fig. 4.4
Approximation quality of answering top-
k
simple typicality queries.
We apply typicality queries on the NBA 2005-2006 Season Statistics
2
. The data
set contains the technical statistics of 458 NBA players, including 221 guards, 182
forwards and 55 centers, on 16 numerical attributes.
As discussed in Section 2.2.1, we can model the set of NBA players as an uncer-
tain object and each player as an instance of the object. Table 4.4 shows the top-3
most typical players, and some of the attribute values. The results answer Jeff's
question in Section 2.2.1.
To answer Jeff's question in Section 2.2.1, we model the set of guards as the
target uncertain object, and the set of forwards and centers as the other two uncer-
2
http://sports.yahoo.com/nba/stats/.
